研究建立了基于时间序列分解的神经网络模型,能对降雨时间序列挖掘并预测。(1)以桓台县1979-2018年的480组月降雨数据为例,将降雨时间序列分解为趋势项、周期项、突变项与随机项。(2)采用累积距平法、Mann-Kendall趋势分析法、Hurst指...研究建立了基于时间序列分解的神经网络模型,能对降雨时间序列挖掘并预测。(1)以桓台县1979-2018年的480组月降雨数据为例,将降雨时间序列分解为趋势项、周期项、突变项与随机项。(2)采用累积距平法、Mann-Kendall趋势分析法、Hurst指数法、特征点法方法进行趋势性分析;小波分析法进行周期性分析;Mann-Kendall突变检验法和Pettitt法进行突变性分析;采用自相关法和单位根法对随机项进行检验。(3)以1979-2014年的432组月降雨时间序列随机项为率定数据,2015-2016年数据为验证数据,分别建立NAR(Nonlinear Auto Regression)与NARX(Nonlinear Auto Regression with External Input)神经网络随机项预测模型,对2017-2018年月降雨数据进行预测,并与直接预测结果对比。结果表明:(1)桓台县1979-2018年月降雨量数据有微弱的上升趋势,预测未来将呈微弱下降趋势,其第一主周期是19(月),数据不存在明显的突变情况。(2)NAR神经网络所得2017-2018年的月降雨量预测值与实测值误差为16.79%。展开更多
Tracing erosion flux within a single catchment is one of the major targets for the Earth's Critical Zone science. The sedimentary succession in landslide-dammed reservoirs within the Chinese Loess Plateau(CLP) ser...Tracing erosion flux within a single catchment is one of the major targets for the Earth's Critical Zone science. The sedimentary succession in landslide-dammed reservoirs within the Chinese Loess Plateau(CLP) serves as a valuable archive of past erosion history. Deposition couplets and annual freeze–thaw layers were firstly identified for the sedimentary succession of the Jingbian reservoir on the northern CLP with high-resolution XRF core scanning. The deposition couplets in the reservoir since 1963 A.D. were further dated with ^(137) Cs activity. We found consistent one-to-one correspondence between couplet specific sediment yield and storm intensity. The reconstructed soil erosion history highlights the control of storm intensity and frequency on loess erosion on the northern CLP in the past hundreds of years.展开更多
文摘【目的】水文时间序列的非一致性分析可有效提高水文频率分析计算结果的合理性与准确性。非一致性序列的研究对象主要集中于洪水和径流,而对降雨的研究相对较少。【方法】运用Mann-Kendall(MK)趋势检验法以及广义可加模型(Generalized Additive Models for Location Scale and Shape,GAMLSS),以古黄河宿迁段流域的10个雨量站1980—2018年逐月降雨数据为例,研究年、汛期以及非汛期降雨序列的非一致性特征并分析非一致性特征对降雨频率设计值的影响。【结果】结果如下:(1)在5%的显著性水平下,MK法检验的结果表明研究区各降雨序列虽存在趋势但不显著;(2)GAMLSS模型可以揭示MK检验无法反映的显著趋势,从而发现研究区年、汛期、非汛期降雨序列的非一致性特征较为明显,且主要表现为方差非一致性;(3)由于降雨序列的非一致性特征,一致性模型存在高估或低估的风险,且不同时期的频率设计值呈现较大的差异,新袁站非汛期降雨、金锁站年降雨某一时期的2 a一遇降雨量级甚至可能超过其他时期的50 a一遇降雨量级。【结论】应综合考虑时间趋势和频率要素来推求降雨频率设计值,可为GAMLSS模型在流域降雨非一致性定量分析的应用中提供参考。
文摘研究建立了基于时间序列分解的神经网络模型,能对降雨时间序列挖掘并预测。(1)以桓台县1979-2018年的480组月降雨数据为例,将降雨时间序列分解为趋势项、周期项、突变项与随机项。(2)采用累积距平法、Mann-Kendall趋势分析法、Hurst指数法、特征点法方法进行趋势性分析;小波分析法进行周期性分析;Mann-Kendall突变检验法和Pettitt法进行突变性分析;采用自相关法和单位根法对随机项进行检验。(3)以1979-2014年的432组月降雨时间序列随机项为率定数据,2015-2016年数据为验证数据,分别建立NAR(Nonlinear Auto Regression)与NARX(Nonlinear Auto Regression with External Input)神经网络随机项预测模型,对2017-2018年月降雨数据进行预测,并与直接预测结果对比。结果表明:(1)桓台县1979-2018年月降雨量数据有微弱的上升趋势,预测未来将呈微弱下降趋势,其第一主周期是19(月),数据不存在明显的突变情况。(2)NAR神经网络所得2017-2018年的月降雨量预测值与实测值误差为16.79%。
基金financially supported by the 973Program(No.2013CB956402)National Natural Science Foundation of China(No.41225015)
文摘Tracing erosion flux within a single catchment is one of the major targets for the Earth's Critical Zone science. The sedimentary succession in landslide-dammed reservoirs within the Chinese Loess Plateau(CLP) serves as a valuable archive of past erosion history. Deposition couplets and annual freeze–thaw layers were firstly identified for the sedimentary succession of the Jingbian reservoir on the northern CLP with high-resolution XRF core scanning. The deposition couplets in the reservoir since 1963 A.D. were further dated with ^(137) Cs activity. We found consistent one-to-one correspondence between couplet specific sediment yield and storm intensity. The reconstructed soil erosion history highlights the control of storm intensity and frequency on loess erosion on the northern CLP in the past hundreds of years.
